They rarely see much use, though, because most of us quickly experience that the type annotations that seem essential when we work on statically typed languages quickly become less essential once you adopt a more idiomatic Ruby style.

E.g. if you want to output something as text, rather than dictate what should be passed in, call #to_s on it (and optionally handle the failure if it doesn't implement #to_s, if it is reasonable to continue).

Once you adopt the attitude of asking for what you actualy need rather than demanding the client pass in what you think should be passed in, type annotations start feeling like a clamp around your foot rather than a necessity in most cases, and most of the remaining assistance they could provide tends to fall away with test cases you would require in either case.

I used to be a static typing zealot, and I still want things to be "as static as possible", but I've come to accept that most of the time the burdens it adds are not worth the benefits.

I'm sure we can do better, and there are certainly cases where optional static type annotation could be helpful, but at this point I'm not giving up the expressiveness dynamc typing gives me - a static system would need to be practically "invisible" for me to find it acceptable.

I won't deny there are some clever things one can do to improve code with sophisticated typing. However, the typical usage is to alleviate the mistakes of bad names and poor design. When I find type errors at runtime, I look for a way to refactor that avoids confusion without needing to add type checking. I don't always succeed, but when I do the code is more elegant.